More on the World of Biological Databases

Today is our day of frustration. It started with an unsuccessful attempt to get raw reads for an already published paper. We were debating about the difficulty of some calculations on that huge RNAseq data set, but never expected retrieving the libraries to be the biggest hurdle.

Anyway, we decided to move on to a different analysis that also seemed very easy at first. Let us explain what the plan was and where we stumbled, because it is a good example of the messy world of biological databases that we discussed previously.

Here was the planned analysis. We wanted to find out the genes neighboring a set of microRNAs in the genome of mosquito Aedes aegypti. Very easy to do it from annotation files, right?

Why Aedes aegypti? For no good reason other than its name starting with A. Let us explain. For those unfamiliar with miRNAs, miRBase is the central repository for miRNAs from all organisms. The database has a very helpful ftp archive, where it splits annotations for various organisms. Aedes aegypti happens to be the first one in that list.

We picked the aae.gff3 from miRbase, and thought the rest of the tasks would be simple. The gff3 file gave the coordinates for genome assembly AaegL1. So, we thought it would be a matter of few more keystrokes to retrieve the remaining annotations for AaegL1 and find neighboring genes for all miRNAs.

Genomic data for Aades aegypti is hosted by VectorBase - a database maintained by Harvard, Ensembl and other very smart groups of people. The latest annotation file I found there was for A. aegypti annotation GFF, Aaeg1.3 geneset. It was not clear, whether the annotation was for genome assembly AaegL1, but given that AaegL1 was the latest genome assembly, we presume the annotation file was for the same assembly as the miRNA file in miRBase was.

That should be easy to check by comparing few coordinates from two gff files, right? We failed.

The coordinates in miRBase file are given in terms of contigs.

The coordinates in Vectorbase file are given in terms of supercontigs.

We could not compare. Where is the contig-supercontig mapping file?

The miRNAs in miRBase were named as miR-124, mir-307, etc. It is the same convention used in papers.

The miRNAs in VectorBase were named as AAEL012345, AAEL123455, etc. Who invented those names, and how do they connect to miR-124, etc.?

At this point, we gave up and went for calculations from first principles. We downloaded the miRNA sequences, downloaded the genome and mapped all genes.